On February 8, 2019 18:39, Junio C Hamano wrote:
> randall.s.becker@xxxxxxxxxx writes:
> > Replaced subtest 15 (CONTENT_LENGTH overflow ssite_t) use of /dev/zero
> > with yes and a translation of its result to a stream of NULL. This is
> > a more portable solution.
> >
> > Signed-off-by: Randall S. Becker <rsbecker@xxxxxxxxxxxxx>
> > ---
> > t/t5562-http-backend-content-length.sh | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/t/t5562-http-backend-content-length.sh
> > b/t/t5562-http-backend-content-length.sh
> > index 90d890d02..b8d1913e5 100755
> > --- a/t/t5562-http-backend-content-length.sh
> > +++ b/t/t5562-http-backend-content-length.sh
> > @@ -143,14 +143,14 @@ test_expect_success GZIP 'push gzipped empty' '
> >
> > test_expect_success 'CONTENT_LENGTH overflow ssite_t' '
> > NOT_FIT_IN_SSIZE=$(ssize_b100dots) &&
> > - env \
> > + yes | tr "y" "\\0" | env \
>
> I do not quite get this use of tr. The original feeds a stream of NULs
out of
> /dev/zero to the command; the yes-to-tr pipe instead feeds a stream of
> alternating NUL and LF.
That's why we're going to go with a generate_zero_bytes function per Peff.
I'm working on a more comprehensive patch covering t5562, t5318, and
test-lib-functions.sh that will (hopefully) be satisfactory and remove the
dependency on /dev/zero and fixes the related new breakages in 2.21.0-rc0.
The test case in t5318 is specific about wanting zero bytes although the
test is just intending to generate a corrupt block that generates a
different hash, so yes 'yes' might be sufficient, but I don't like
randomness myself if we're taking two different tests being involved.
My current intent is to add to test-lib-functions.sh, a method of
generalizing blocks of zeros to a pipe:
+# Generate an output of $1 bytes of all zeroes (NULs, not ASCII zeroes).
+# If $1 is < 0, output forever or until the receiving pipe stops reading,
whichever comes first.
+ generate_zero_bytes () {
+ perl -e ' if ($ARGV[0] < 0) { while (-1) { print "\0" } } else {
print "\0" x $ARGV[0] }' "$@"
+ }
And then fit that into the two tests, then submit as a patch.
> Does the actual bytes fed to the consumer make any difference? If not,
> perhaps we can use 'yes' as-is?
>
> > CONTENT_TYPE=application/x-git-upload-pack-request \
> > QUERY_STRING=/repo.git/git-upload-pack \
> > PATH_TRANSLATED="$PWD"/.git/git-upload-pack \
> > GIT_HTTP_EXPORT_ALL=TRUE \
> > REQUEST_METHOD=POST \
> > CONTENT_LENGTH="$NOT_FIT_IN_SSIZE" \
> > - git http-backend </dev/zero >/dev/null 2>err &&
> > + git http-backend >/dev/null 2>err &&
> > grep "fatal:.*CONTENT_LENGTH" err
> > '
Regards,
Randall
